Complex system health analysis by the Graphical Evolutionary Hybrid Neuro-Observer (GNeuroObs)

Abstract

Obtaining methodologies that enable predictive health monitoring of components degradation and the propagation of related effects across the overall system is a need when designing complex systems (such as autonomous vehicles, robotic systems, and aerospace platforms). In this paper, a current software development is presented for workflow generation and visualization to evaluate how component degradation impacts an entire system. Relevant technical aspects of this “Graphical Evolutionary Hybrid Neuro-Observer” (GNeuroObs) include: (a) highly accurate system modeling; (b) techniques for system level analysis; and (c) low level entity instantiations that builds on health monitoring and root cause analysis. The GNeuroObs is described with the application of a fuel subsystem. In that system, the methodology allows for describing interrelations among a set of heterogeneous sensors, where Health Monitoring algorithms are used to analyze failures in entities and propagation of effects across the system.